Starshot: all tested images returns the same result.

[effelle]

Describe the bug We are trying to implement pylinac in our workflow but sadly we were unable to archieve solid results (and I'm pretty sure is our fault!).

To Reproduce Using the following script form docs:

from turtle import clear
from scipy import ndimage

import pylinac
from pylinac.core.image_generator import GaussianFilterLayer, FilteredFieldLayer, AS1200Image, RandomNoiseLayer

star_path = 'C:/Users/effel/Downloads/radioterapia/Images/CR.1.3.51.0.7.2652200017.39611.53065.42883.31209.9902.52590.dcm'
as1200 = AS1200Image()
for _ in range(6):
    as1200.add_layer(FilteredFieldLayer((270, 5), alpha=0.5))
    as1200.image = ndimage.rotate(as1200.image, 30, reshape=False, mode='nearest')
as1200.add_layer(GaussianFilterLayer(sigma_mm=3))
as1200.generate_dicom(file_out_name=star_path)

# analyze it
star = pylinac.Starshot(star_path)  
star.analyze()
print(star.results())
star.plot_analyzed_image()

Result: PASS 

The minimum circle that touches all the star lines has a diameter of 0.045 mm. 

The center of the minimum circle is at 639.5, 639.5

Expected behavior The analyzed image should return different values

Screenshots

Additional context I'm quite sure we are missing something since I didn't find similar open issues for the starshot. Plese note the script always returns some warnings with our images:

C:\Users\effel\AppData\Local\Programs\Python\Python310\lib\site-packages\pydicom\valuerep.py:290: UserWarning: The value length (36) exceeds the maximum length of 16 allowed for VR SH.
  warnings.warn(msg)
C:\Users\effel\AppData\Local\Programs\Python\Python310\lib\site-packages\pydicom\valuerep.py:290: UserWarning: The value length (18) exceeds the maximum length of 16 allowed for VR SH.
  warnings.warn(msg)

I can provide the Dicom files used for the test if needed.
Thank you in advance for you help!

F.

[jrkerns]

F, I don't know if you are trying to analyze your own images or perform benchmarking as your scripts show.

Just to clarify, your scripts are generating artificial images and processing them. I can't tell any difference in code between your screenshots; maybe I missed it. You can use a gist to dump code into.

If you are trying to process your OWN images then drop the image generator part of the script and just follow the typical use pattern: https://pylinac.readthedocs.io/en/release-3.0/starshot_docs.html#typical-use

If you are trying to produce artificial images then you shouldn't expect differences between them.

In the future, you can post on the pylinac google group for discussions like this.

[effelle]

Hi James, the first screenshot is there just to show how the beaming stars are different between the images. The other 3 following screenshot show how the result is the same even if the input image is not (name of the file). I do apologize if was not clear. If I get it well I shuld use this script instead:

from pylinac import Starshot
star_img = 'C:/Users/effel/Downloads/radioterapia/Images/CR.1.3.51.0.7.2652200017.39611.53065.42883.31209.9902.52590.dcm'
mystar = Starshot(star_img)
mystar.analyze(radius=0.5, tolerance=0.8)
print(mystar.results())
mystar.plot_analyzed_image()

But the result is always the same:

Result: PASS 

The minimum circle that touches all the star lines has a diameter of 0.020 mm. 

The center of the minimum circle is at 639.4, 639.5

As the image produced:

I'm starting to ask myself if I can use a Dicom image as input or if I must use a TIF format.

F.

[effelle]

So after some trials and errors I found what it seems to work for us:

from pylinac import Starshot
img_path = 'C:/Users/effel/Downloads/radioterapia/Images/COLTESTE02.tif'
mystar = Starshot(img_path, dpi=105, sid=1000)
mystar.analyze()
print(mystar.results())
mystar.plot_analyzed_image()

Using tif files, not Dicom.